Search papers, labs, and topics across Lattice.
The paper introduces Parcae, a stable looped language model architecture that addresses instability issues in existing looped models by constraining the spectral norm of injection parameters via discretization of a negative diagonal parameterization. This approach allows for predictable scaling of FLOPs during training and test-time while keeping the parameter count fixed, leading to improved validation perplexity and quality compared to prior looped models and Transformer baselines. Scaling experiments up to 1.3B parameters demonstrate that Parcae achieves up to 87.5% of the quality of a Transformer twice its size under a fixed parameter and data budget.
Looped language models can now rival Transformers in quality at a fraction of the parameter count, thanks to a new architecture that tames their notorious instability.
Traditional fixed-depth architectures scale quality by increasing training FLOPs, typically through increased parameterization, at the expense of a higher memory footprint, or data. A potential alternative is looped architectures, which instead increase FLOPs by sending activations through a block of layers in a loop. While promising, existing recipes for training looped architectures can be unstable, suffering from residual explosion and loss spikes. We address these challenges by recasting looping as a nonlinear time-variant dynamical system over the residual stream. Via a linear approximation to this system, we find that instability occurs in existing looped architectures as a result of large spectral norms in their injection parameters. To address these instability issues, we propose Parcae, a novel stable, looped architecture that constrains the spectral norm of the injection parameters via discretization of a negative diagonal parameterization. As a result, Parcae achieves up to 6.3% lower validation perplexity over prior large-scale looped models. Using our stable looped architecture, we investigate the scaling properties of looping as a medium to improve quality by increasing FLOPs in training and test-time. For training, we derive predictable power laws to scale FLOPs while keeping parameter count fixed. Our initial scaling laws suggest that looping and data should be increased in tandem, given a fixed FLOP budget. At test-time, we find that Parcae can use looping to scale compute, following a predictable, saturating exponential decay. When scaled up to 1.3B parameters, we find that Parcae improves CORE and Core-Extended quality by 2.99 and 1.18 points when compared to strong Transformer baselines under a fixed parameter and data budget, achieving a relative quality of up to 87.5% a Transformer twice the size.